A Person Re-identification Data Augmentation Method with Adversarial Defense Effect

Gong, Yunpeng; Zeng, Zhiyong; Chen, Liwen; Luo, Yifan; Weng, Bin; Ye, Feng

doi:10.48550/arxiv.2101.08783

Cited by 2 publications

(7 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast, our method maintains the rank-1 accuracy at 6.3% on Market-1501 even when attacked by Mis-rank [50], the most powerful attacker of re-ID models. For re-ID systems strengthened by defense methods such as LGPR [16], GGPR [16] and MMD [16], we can observe improved robustness under the two attacks. For example, HACNN [30] sees its rank-1 value increase from 0.9 to 3.1 if HACNN [30] is defended by MMD [16].…”

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 89%

“…In this experiment, we conduct white-box attack experiments to evaluate the robustness of our method. We also compared our method with the state-of-the-art (SOTA) re-ID models combined with recent defense methods, i.e., Gray Patch Replacement (GPR) [16], Multi-modal defense (MDD) [16] and AMA [5]. GPR [16] is an augmentation method that leverages the information between the color and the spatial structure.…”

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

“…We also compared our method with the state-of-the-art (SOTA) re-ID models combined with recent defense methods, i.e., Gray Patch Replacement (GPR) [16], Multi-modal defense (MDD) [16] and AMA [5]. GPR [16] is an augmentation method that leverages the information between the color and the spatial structure. GPR [16] can change the input image with both global and local grayscale patch replacement (known as GGPR and LGPR) as shown in Fig.…”

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

“…GPR [16] is an augmentation method that leverages the information between the color and the spatial structure. GPR [16] can change the input image with both global and local grayscale patch replacement (known as GGPR and LGPR) as shown in Fig. 4.…”

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

“…4. Multi-modal defense (MDD) improves GPR [16] by integrating three image modalities: visible, grayscale and sketch. The adversarial metric attack (AMA) [5] is a SOTA adversarial metric defense for re-ID that attacks the metrics used to train a metric-preserving network.…”

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

See 4 more Smart Citations

Generative Metric Learning for Adversarially Robust Open-world Person Re-Identification

Liu

Hong

et al. 2023

ACM Trans. Multimedia Comput. Commun. Appl.

View full text Add to dashboard Cite

The vulnerability of re-identification (re-ID) models under adversarial attacks is of significant concern as criminals may use adversarial perturbations to evade surveillance systems. Unlike a closed-world re-ID setting (i.e., a fixed number of training categories), a reliable re-ID system in the open-world raises the concern of training a robust yet discriminative classifier, which still shows robustness in the context of unknown examples of an identity. In this work, we improve the robustness of open-world re-ID models by proposing a generative metric learning approach to generate adversarial examples that are regularized to produce robust distance metric. The proposed approach leverages the expressive capability of generative adversarial networks to defend the re-ID models against feature disturbance attacks. By generating the target people variants and sampling the triplet units for metric learning, our learned distance metrics are regulated to produce accurate predictions in the feature metric space. Experimental results on the three re-ID datasets, i.e., Market-1501, DukeMTMC-reID and MSMT17 demonstrate the robustness of our method.

show abstract

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 89%

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

Section: Uantitative Analysis Under Adversarial Atacksmentioning

confidence: 99%

See 3 more Smart Citations

Generative Metric Learning for Adversarially Robust Open-world Person Re-Identification

Liu

Hong

et al. 2023

ACM Trans. Multimedia Comput. Commun. Appl.

View full text Add to dashboard Cite

show abstract

PatchMix Augmentation to Identify Causal Features in Few-Shot Learning

Liu

Sun

et al. 2023

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

The task of Few-shot learning (FSL) aims to transfer the knowledge learned from base categories with sufficient labelled data to novel categories with scarce known information. It is currently an important research question and has great practical values in the real-world applications. Despite extensive previous efforts are made on few-shot learning tasks, we emphasize that most existing methods did not take into account the distributional shift caused by sample selection bias in the FSL scenario. Such a selection bias can induce spurious correlation between the semantic causal features, that are causally and semantically related to the class label, and the other non-causal features. Critically, the former ones should be invariant across changes in distributions, highly related to the classes of interest, and thus well generalizable to novel classes, while the latter ones are not stable to changes in the distribution. To resolve this problem, we propose a novel data augmentation strategy dubbed as PatchMix that can break this spurious dependency by replacing the patch-level information and supervision of the query images with random gallery images from different classes from the query ones. We theoretically show that such an augmentation mechanism, different from existing ones, is able to identify the causal features. To further make these features to be discriminative enough for classification, we propose Correlation-guided Reconstruction (CGR) and Hardness-Aware module for instance discrimination and easier discrimination between similar classes. Moreover, such a framework can be adapted to the unsupervised FSL scenario. The utility of our method is demonstrated on the state-of-the-art results consistently achieved on several benchmarks including miniImageNet, tieredImageNet, CIFAR-FS , CUB, Cars, Places and Plantae, in all settings of single-domain, cross-domain and unsupervised FSL. By studying the intra-variance property of learned features and visualizing the learned features, we further quantitatively and qualitatively show that such a promising result is due to the effectiveness in learning causal features.

show abstract

A Person Re-identification Data Augmentation Method with Adversarial Defense Effect

Cited by 2 publications

References 24 publications

Generative Metric Learning for Adversarially Robust Open-world Person Re-Identification

Generative Metric Learning for Adversarially Robust Open-world Person Re-Identification

PatchMix Augmentation to Identify Causal Features in Few-Shot Learning

Contact Info

Product

Resources

About